[patch] This patch fixes the issue by using matrix addressing instead (whose cost is negligible anyways).

this.patch (text/plain), 1.76 KB, created by Francisco Blas Izquierdo Riera on 2013-05-29 23:00:38 UTC

--- a.c	2013-05-30 00:47:51.100621173 +0200
+++ b.c	2013-05-30 00:56:27.318621998 +0200
@@ -38,12 +38,6 @@
     "imul                 %%rdx, %%rax\n\t" \
     "add                  %%rax, %%rsi\n\t"
 
-#define FIRMULREG(label, offset, firc)\
-    LABEL_MANGLE(label)":       \n\t" \
-    "movslq "#offset"(%0), %%rax\n\t" \
-    "imul        %"#firc", %%rax\n\t" \
-    "add            %%rax, %%rsi\n\t"
-
 #define CLEAR_ACCUM                   \
     "xor            %%rsi, %%rsi\n\t"
 
@@ -63,9 +57,6 @@
     "add                %%eax , %%esi\n\t" \
     "adc                %%edx , %%ecx\n\t"
 
-#define FIRMULREG(label, offset, firc)  \
-    MLPMUL(label, #offset, "0", "0")
-
 #define CLEAR_ACCUM                  \
     "xor           %%esi, %%esi\n\t" \
     "xor           %%ecx, %%ecx\n\t"
@@ -108,9 +99,9 @@
         FIRMUL   (ff_mlp_firorder_6, 0x14   )
         FIRMUL   (ff_mlp_firorder_5, 0x10   )
         FIRMUL   (ff_mlp_firorder_4, 0x0c   )
-        FIRMULREG(ff_mlp_firorder_3, 0x08,10)
-        FIRMULREG(ff_mlp_firorder_2, 0x04, 9)
-        FIRMULREG(ff_mlp_firorder_1, 0x00, 8)
+        FIRMUL   (ff_mlp_firorder_3, 0x08   )
+        FIRMUL   (ff_mlp_firorder_2, 0x04   )
+        FIRMUL   (ff_mlp_firorder_1, 0x00   )
         LABEL_MANGLE(ff_mlp_firorder_0)":\n\t"
         "jmp  *%6                     \n\t"
         IIRMUL   (ff_mlp_iirorder_4, 0x0c   )
@@ -137,9 +128,6 @@
           /* 3*/"+r"(blocksize)
         : /* 4*/"r"((x86_reg)mask), /* 5*/"r"(firjump),
           /* 6*/"r"(iirjump)      , /* 7*/"c"(filter_shift)
-        , /* 8*/"r"((int64_t)coeff[0])
-        , /* 9*/"r"((int64_t)coeff[1])
-        , /*10*/"r"((int64_t)coeff[2])
         : "rax", "rdx", "rsi"
 #else /* ARCH_X86_32 */
           /* 3*/"+m"(blocksize)